LAION-5B is an open dataset of about 5.85 billion image links paired with their text captions, collected from the web and used to train image models such as Stable Diffusion.
In plain words
Models like CLIP learn from huge numbers of image-caption pairs, but such datasets were private. LAION released LAION-5B in 2022: web image links plus captions, filtered with CLIP, with 2.32 billion in English. It stores links, not the images. In December 2023 a Stanford Internet Observatory report found links to child sexual abuse material, so LAION withdrew it. A cleaned version, Re-LAION-5B, came out in August 2024.
A quick example
A researcher replicating an image-text model would download the dataset's list of links and captions, fetch the images still online, and train on those pairs.
Where you'll meet it
- Training and studying open image-text and text-to-image models.
- Research on web-scale dataset safety and curation.
Related
Sources
- PaperLAION-5B: An open large-scale dataset for training next generation image-text models, Schuhmann et al. (arXiv) · read 29 Sept 2026
- DocsReleasing Re-LAION 5B, LAION · read 26 Oct 2026
Checked by WhatIsAI against the sources above · 26 Oct 2026. Spotted an error? Report it.